Virtualized ANSI C Macros for Finite Volume and Finite Element Simulations <author> <version> <abstract> These macro calls constitute a package which has been specifically designed to describe CFD algorithms in a data-parallel fashion. The source program may be translated to serial ANSI C, vectorized C, C*, MasPar's MPL/C, or ORNL's PVM/C. The source program is free from many of the typical dependencies on the target architecture. These macros address the problem of machine independent parallelization of computational fluid dynamics codes on a number of commercially available parallel computers. Machine independence was achieved by using a macro package that augments the C language with a set of software constructs to express the parallel algorithm in a data parallel fashion. Efficiency was achieved by separating out the computation parts from the communication parts and by grouping of individual data transfers into large messages. The latter enhancement was used on architectures where message startup latency is high. <description><url>http://sdcd.gsfc.nasa.gov/ESS/exchange/contrib/mobarry/vic.html </url> <reference><url>ftp://sdcd.gsfc.nasa.gov/pub/HPCC/ESS/mobarry/vic.ps</url> <contact>Clark M. Mobarry / k6cmm@supernova.gsfc.nasa.gov <keywords>parallel fluid simulation; computational fluid dynamics; data parallelism; architecture independence; parallel application library <category>application <publication-date> <environment> <method> <application>parallel fluid simulation, computational fluid dynamics <comments> </urc>

ftp://sdcd.gsfc.nasa.gov/pub/HPCC/ESS/mobarry/vic.tar.Z Virtualized ANSI C Macros for Finite Volume and Finite Element Simulations <author> <version> <abstract> These macro calls constitute a package which has been specifically designed to describe CFD algorithms in a data-parallel fashion. The source program may be translated to serial ANSI C, vectorized C, C*, MasPar's MPL/C, or ORNL's PVM/C. The source program is free from many of the typical dependencies on the target architecture. These macros address the problem of machine independent parallelization of computational fluid dynamics codes on a number of commercially available parallel computers. Machine independence was achieved by using a macro package that augments the C language with a set of software constructs to express the parallel algorithm in a data parallel fashion. Efficiency was achieved by separating out the computation parts from the communication parts and by grouping of individual data transfers into large messages. The latter enhancement was used on architectures where message startup latency is high. <description><url>http://sdcd.gsfc.nasa.gov/ESS/exchange/contrib/mobarry/vic.html </url> <reference><url>ftp://sdcd.gsfc.nasa.gov/pub/HPCC/ESS/mobarry/vic.ps</url> <contact>Clark M. Mobarry / k6cmm@supernova.gsfc.nasa.gov <keywords>parallel fluid simulation; computational fluid dynamics; data parallelism; architecture independence; parallel application library <category>application <publication-date> <environment> <method> <application>parallel fluid simulation, computational fluid dynamics <comments> </urc>